As it stands, most industrial companies do not react to breakdowns until their machines and systems have already come to a standstill (reactive maintenance). In some other cases, they conduct maintenance at scheduled intervals (preventative maintenance). However, the former approach can lead to short-term production downtimes, while the latter can result in wasting of machine working hours and material, because components are replaced prematurely. Now as ever, one of the biggest challenges for manufacturing companies is to determine the optimal time for conducting machine maintenance. Data-driven models of production systems are becoming increasingly important in this context. However, since they only extrapolate trends based on past events, there is no data on breakdown scenarios that have not yet occurred. A promising approach for predictive maintenance is to simulate these kinds of breakdown scenarios using physics-based modeling and then apply the data generated in the simulations to enhance data-driven forecast models. What makes this approach possible is a digital twin of the production system, which can be used to make predictive statements about such things as the remaining useful life (RUL) of machine components.
As part of the EU project Z-BRE4K, the Fraunhofer Institute for Industrial Engineering IAO, in cooperation with the Institute of Human Factors and Technology Management IAT from the University of Stuttgart and a further 17 partners from nine European countries, developed concepts and solutions for data acquisition, processing and analysis based on digital twins. The goal of the three-year project was to optimize production systems by means of predictive maintenance in order to save costs and increase companies’ productivity. Joachim Lentes, Team Leader for Digital Engineering at Fraunhofer IAO, explains yet another benefit: “Susceptibility to breakdowns and uncertainty regarding the predictability of resources and capacities are further exacerbated by supplier dependence and unanticipated bottlenecks, especially in times of crisis. That became clear over this past year. Intelligent management of resources and capacities can help companies improve their resilience, as they then gain stability and certainty in their planning. This is precisely what makes it possible to use a digital twin for predictive maintenance.”
Using internationally-tested solutions to create a digital twin for various production environments
The research team defines a digital twin as a digital representation of an actual product (e.g. a machine) or process that includes all data and information relevant to the respective use case throughout its entire life cycle. The basic digital model of the actual object is continually enhanced through operating data, behavior analysis and simulations. Among other things, this process can be used to extrapolate the machinery’s RUL. This approach to RUL estimation is based on statistical methods and algorithms derived from those methods, and consequently, it requires large quantities of data. In this project, the research team from Fraunhofer IAO and the IAT at the University of Stuttgart was responsible in particular for the physics-based modeling. This was needed to conduct simulations of machine behavior, which serve as the basis for the digital twins of various production systems. Once developed, these concepts and solutions were tested, validated and applied in three pilot cases with industrial users, including PHILIPS, the Dutch provider in the field of personal health and health systems, the Spanish supplier GESTAMP and the Italian companies SACMI, a machine manufacturer, and CDS, a manufacturer of plastic closures.
Implementing a predictive maintenance strategy requires a structured approach
The research team has continually put the results of the Z-BRE4K project to use in industry projects and innovation networks. In the next step, having been tested internationally and in various sectors, these solutions are to be transferred to the German industrial landscape to a greater extent. The experience of the research team attests that companies generally do not have an overview of the existing data and corresponding modules that are required for implementing an integrated predictive maintenance strategy. In addition, companies often lack a systematic approach required to establish a business model for implementing predictive maintenance into their manufacturing settings. Two foundational elements are needed to integrate predictive maintenance into existing company processes: an industrial internet-of-things architecture, and a systematic approach derived from this architecture. The experts at Fraunhofer IAO support companies at every stage of this endeavor, from creating tailored road maps and offering neutral, independent advice on using the necessary technologies and software, right through to implementing concrete solutions. “The full potential of a digital twin can only be realized by using all essential data and information in an integrated manner across the entire life cycle,” says Andreas Werner of Fraunhofer IAO. “Speaking from a purely technical perspective, there is already a lot that can be achieved with a digital twin. However, our experience shows that a lot of underlying work is still to be conducted at companies, especially with regards to the organizational aspects.”